Planners should spend the week deciding, not collecting
Module · Planning and Performance
Eight of the nine weeks in a planning cycle go on collection, chasing and re-cutting.
Prophesee FP&A takes that work off the team, reconciles the target and the build by arithmetic rather than by argument, and gives every analyst the means to test an idea and prove it without asking anyone.
The budget takes nine weeks. Eight of them are plumbing.
Collection, chasing, version control, re-cutting. The analytical work, the part the team was actually hired for, gets whatever is left of week nine. The target and the build are settled in a room because no mechanism has ever existed to settle them any other way.
Sources: Association for Financial Professionals, FP&A Benchmarking Survey: Integrated Planning, 2026 (n=332 FP&A and finance professionals across 54 countries) · Association for Financial Professionals, FP&A Benchmarking Survey, May 2025, on planning platform and spreadsheet use · Gartner, 2026 CFO Agenda research, fielded Aug 2025, published Dec 2025 (200+ CFOs) · 3RDi review of published planning platform product pages and product documentation, Aug 2026.
From negotiating the number to reconciling it
ForesightThe forecast is last year's actuals with a growth rate on top, delivered as one number with no band and no driver. Every reviewer discounts it by their own margin.
Gradient-boosted trees and time series models per line, entity and driver, each projection carrying a confidence interval and ranked drivers. Causality tests which operational series lead the financial one, so a driver is measured, not asserted. Monitored for drift and retrained.
PulseThe plan is tested against actuals once a month, in a meeting, weeks after the divergence began. Nothing tests the forecast itself, so a miss is reported rather than prevented.
Write the tolerance in plain English: this cost centre, this margin floor, this hiring ramp. Rules run against actuals and the forward projection, so the alert fires when the plan is projected to miss, not at month close. Scored and routed with the action attached.
HorizonScenarios are built once, presented once and never opened again. Nobody tracks whether the chosen one happened, so no plan is ever scored and no decision is ever attributed.
Predict the no-action trajectory, model the intervention and watch the projected curve update as the assumption changes. Then track actuals against that plan, with counterfactual modelling separating the effect of the action from what the market would have done anyway.
NexusEach ledger carries its own account structure and each entity its own cost centre tree, so the same line means different things in different places.
Ledgers, planning models, headcount systems and written commentary resolved into one graph, with account and cost centre mappings maintained by classification models rather than by hand. Ask in plain language and get the figure with its version, its assumption and the person who owned it.
Turning planning challenges into decisions
The board sets eight percent. The build says four. The gap closes in a room, by seniority, and reappears as a buffer nobody can find.
ForesightOne number, carried forward with a growth rate, with no band and no stated driver. So every layer discounts it and adds its own margin.
Foresight NexusThe number moved and four people spend three days arguing whether it was price, volume, mix or currency before anyone can act.
Foresight PulseThree scenarios are built for the board and never opened again. Nothing records which one was chosen or whether it happened.
HorizonA new account opens in one ledger and lands in Other for two months, so the group view is quietly wrong and nobody is told.
Nexus Pulse12 AI applications that could be relevant
A sample of what becomes possible on the decision layer, not a fixed list: each application draws on the same data foundation and audit trail, and new ones are configured on the engines, not built from scratch.
Top-down target and bottom-up build reconciled by arithmetic, with the gap named, sized and owned.
Every line projected with a confidence band and a ranked list of the drivers behind it.
Plan to actual decomposed into price, volume, mix, rate and currency before the review, not during it.
Where the plan carries hidden contingency, by owner and by line, measured against outturn.
The lines that have begun to drift from plan, ranked by projected impact at year end.
Tolerances written in plain English and tested continuously against actuals and the forward projection.
Account, cost centre and entity mappings maintained continuously, with every drift flagged.
Model the intervention and watch the projected curve update against the target.
Actuals tracked against the modelled plan, with the counterfactual separating the action from the trend.
The date the current plan reaches the target, updated as actuals come in.
Ask the plan a question in plain language and get the figure with its version and its assumption.
Every assumption behind every version, with its owner, its date and what it changed.
A day when the analyst gets to think
Today: A forecast built from twenty submissions, each carrying a buffer nobody can see.
ForesightTarget against build, reconciled by arithmetic. The gap has an owner and a number before the meeting starts.
Today: A line that started drifting in week three, discussed in the week nine review.
PulseTwo cost centres are projected to breach plan by the third quarter. Severity scored, routed, with the variance already decomposed.
Today: Three scenarios built for the board, then never opened again.
HorizonA pricing move tested against a hiring pause on the projected curve. She takes the one that closes the gap without the attrition.
Today: The plan changed. Nobody can say whether the change worked.
HorizonLast quarter's intervention tracked against the modelled plan, with the counterfactual separating it from the market. That goes to the board.
Tomorrow's plan becomes today's decision.
Improve forecast accuracy
We agree the metric and the baseline in week one, and measure the result on your data.